Surface Specification Solutions Sought

• Sheet tracing– Identifiable as local maxima/minima, not boundary

– Folded fabric

– Nasal turbinates

• Mostly-closed volumes; Digital endocasts– (Semi?)-automatically finding closures

– Finding surfaces that only exist in the mind of the scientist…

Nasal turbinates

Project with Tim Rowe (UT) and Blaire Van Valkenburgh (UCLA)

Airbag fabric

Airbag Folds

• Phase 1: Find a good separator material– Need to keep plies distinct

when folded together

• Scanning conditions– 420 kV, 1.8 mA

– Field of view 180 mm

– Slice thickness 0.25 mm

– Slice spacing 0.2 mm

– 30s/slice

Airbag folds

Other difficulties

• Most algorithms (i.e. Marching Cubes) trace interfaces, not local maxima/minima– Would provide a “hollow plane”

• Thinness of sheet– Ensures gray levels inconsistent– Fouls up image processing for noise reduction

Airbag folds

• Phase 2: Conversion to CAD– The tricky part

No good threshold value

Airbag folds

• Folded sheet traversal algorithm– Based on local moment of inertia

Algorithm Outline

• Build using facets (triangles in 3D space)– Use voxel centers as vertices

• Find first facet, and then grow out from each edge– Growth is normal to edge surface, and perpendicular to

3rd eigenvector of endpoints• Search cone to find best target (highest gray value)

– Add best facet (highest gray new voxel) at each step– No concave angles between facets– After facet added, see if new voxel can resolve any

other edges

Algorithm in action

Remaining issues

• Speed– Substantial data sets would take hours

• Filling holes• Avoiding ragged edges• ????

– As more data sets are tried, additional complexities may be revealed

– Bag seams?

“Lunch bag” scan

Digital Endocasts

Your mission, if you choose…

Extra credit